Diversity and function of maternal HIV-1-specific antibodies at the time of vertical transmission

Infants of HIV positive mothers can acquire HIV infection by various routes, but even in the absence of antiviral treatment, the majority of these infants do not become infected. There is evidence that maternal antibodies may provide some protection from infection, but gestational maternal antibodies have not yet been characterized in detail. One of the most studied vertically-infected infants is BG505, as the virus from this infant yielded an Envelope protein that was successfully developed as a stable trimer. Here, we isolated and characterized 39 HIV-specific neutralizing monoclonal antibodies (nAbs) from MG505, the mother of BG505, at a time point just prior to vertical transmission. These nAbs belonged to 21 clonal families, employed a variety of VH genes, many were specific for the HIV-1 Env V3 loop, and this V3 specificity correlated with measurable antibody-dependent cellular cytotoxicity (ADCC) activity. The isolated nAbs did not recapitulate the full breadth of heterologous nor autologous virus neutralization by contemporaneous plasma. Notably, we found that the V3-targeting nAb families neutralized one particular maternal Env variant even though all tested variants had low V3 sequence diversity and were measurably bound by these nAbs. None of the nAbs neutralized the BG505 transmitted virus. Furthermore, the MG505 nAb families were found at relatively low frequencies within the maternal B cell repertoire: all less than 0.25% of total IgG sequences. Our findings demonstrate the diversity of HIV-1 nAbs that exist within a single mother, resulting in a collection of antibody specificities that can shape the transmission bottleneck. Importance Mother-to-child-transmission of HIV-1 offers a unique setting in which maternal antibodies both within the mother and passively-transferred to the infant are present at the time of viral exposure. Untreated HIV-exposed human infants are infected at a rate of 30-40%, meaning that some infants do not get infected despite continued exposure to virus. Since the potential of HIV-specific immune responses to provide protection against HIV is a central goal of HIV vaccine design, understanding the nature of maternal antibodies may provide insights into immune mechanisms of protection. In this study, we isolated and characterized HIV-specific antibodies from the mother of an infant whose transmitted virus has been well studied.


26
Infants of HIV positive mothers can acquire HIV infection by various routes, but even in 27 the absence of antiviral treatment, the majority of these infants do not become infected. There is 28 evidence that maternal antibodies may provide some protection from infection, but gestational 29 maternal antibodies have not yet been characterized in detail. One of the most studied 30 vertically-infected infants is BG505, as the virus from this infant yielded an Envelope protein that 31 was successfully developed as a stable trimer. Here, we isolated and characterized 39 HIV-32 specific neutralizing monoclonal antibodies (nAbs) from MG505, the mother of BG505, at a time 33 point just prior to vertical transmission. These nAbs belonged to 21 clonal families, employed a 34 variety of VH genes, many were specific for the HIV-1 Env V3 loop, and this V3 specificity 35 correlated with measurable antibody-dependent cellular cytotoxicity (ADCC) activity. The 36 isolated nAbs did not recapitulate the full breadth of heterologous nor autologous virus 37 neutralization by contemporaneous plasma. Notably, we found that the V3-targeting nAb 38 families neutralized one particular maternal Env variant even though all tested variants had low 39 V3 sequence diversity and were measurably bound by these nAbs. None of the nAbs 40 neutralized the BG505 transmitted virus. Furthermore, the MG505 nAb families were found at 41 relatively low frequencies within the maternal B cell repertoire: all less than 0.25% of total IgG 42 sequences. Our findings demonstrate the diversity of HIV-1 nAbs that exist within a single 43 mother, resulting in a collection of antibody specificities that can shape the transmission

186
To map epitope specificities of each clonal nAb family, we used a combination of phage

198
MG505.52), we observed weak but significant enrichments of a peptide that truncated the 199 minimal epitope sequence suggesting this is a core part of the epitope ( Figure 3A, blue 200 residues).

201
Since the library included sequences of several different HIV-1 variants, our PhIP-seq 202 data also allowed us to gain some insight into which amino acids were preferred at highly 203 variable residues within the library sequences. Interestingly, while we observed some variation 3B). In all nine cases, the V3-peptides that bound to the nAbs were identical to the 213 BG505.W6.C2 Env V3 sequence (Supplementary Figure 1), which is also identical to the V3 214 regions of six of seven MG505 W0 Env sequences ( Figure 3B and (1)).

219
The remaining 12 nAb family representatives that were tested by PhIP-seq did not 220 significantly enrich for any phage in the library, suggesting that they may target conformational 221 epitopes that cannot be detected in the 39-mer peptides expressed by phage in the library. To 222 broadly map the epitope specificities of these 12 non-V3-specific nAbs, we employed ELISA 223 assays using gp120 and gp41 antigens. Eleven of twelve families bound gp120 monomer 11

252
The V3 region of MG505.W0.G2 is identical to that of the other maternal and infant 253 viruses, with the exception of MG505.W0.H3 which has a R310H substitution ( Figure 3B). To 254 explore the possibility that the V3 nAbs bound the majority of the MG505 and BG505 variants, 255 even though they did not neutralize them, we tested them for Env binding via cell-surface 256 binding assays. Indeed, detectable binding was observed for all V3 nAbs to all autologous Envs 257 expressed on the surface of cells ( Figure 5B). We also tested three gp120 nAbs that were not 258 V3-specific. These also bound to cell surface-expressed Env, albeit at lower levels ( Figure    290 families 3, 9, and 13 ( Figure 6A). These nAb families ranked 65 th , 400 th , and 278 th largest within 291 the IgG repertoire, representing 0.24%, 0.08%, and 0.13% of the total repertoire, respectively.

292
We did not sample any additional clonal IgG sequences stemming from the other 18 nAb 293 families, indicating that family members of these nAbs either absent or rare enough that we did 294 not achieve sufficient sampling depth to detect them (Table 1).

295
Repertoire analysis is more nuanced for antibody light chains because of the vastly 296 lower theoretical diversity due to the absence of D genes, shorter CDR3 lengths, and shorter

337
The MG505 nAbs were isolated using a high-throughput functional screening method in 338 technical replicate which allowed for less biased discovery of HIV-neutralizing antibodies than 339 Env bait-based approaches, which tend to target antibodies of certain specificities. The diversity 340 of gene usage within the resulting 39 nAbs was striking, especially considering that the screen 341 was not saturating and that there were likely additional families that contributed to the breadth of 342 the MG505 response. Also striking was the fact that we isolated multiple members of 10 clonal 343 families despite their relative rarity in the repertoire, as was indicated by our failure to find

362
We used a novel phage immunoprecipitation approach to map the epitopes of these 363 antibodies and found that 9 of the 21 families recognized a linear epitope in V3 that includes the 364 GPGQ sequence that is a common target for V3 nAbs (28,29). These V3-specific nAbs were

375
The most potent ADCC-mediating nAbs were all V3-specific. The ability to mediate 376 ADCC was correlated with both V3 specificity and the ability to neutralize a Tier 2 virus. It is 377 possible that V3 specificity could simply enable Tier 2 breadth and ADCC function, especially 378 since V3 antibodies can exhibit weak breadth for Tier 2 viruses due to the viruses' sampling of 379 the open conformational states that allow for V3 binding (31). Given that ADDC activity has 380 been correlated with infant outcomes (13), these highly potent maternal ADCC nAbs may 381 provide clues to the mechanisms of protection of ADCC antibodies.

382
It was surprising that the V3 nAbs only neutralized one maternal variant, despite all 383 maternal and infant viruses having identical sequences within the V3 minimal epitope. It is 384 possible that V3 defines only part of the epitope for these nAbs and/or that factors other than 385 minimal epitope binding affect pseudovirus neutralization, such as occlusion of V3 in the native 386 trimer (32). Despite the lack of autologous virus neutralization, we could detect binding to the 387 cell surface-expressed forms of Env of these same viruses. We also detected binding to the 388 BG505 infant Env SOSIP trimer but not neutralization of the corresponding virus. The single 389 difference between the BG505 Envs used in the binding versus neutralization studies is that the 390 native-like SOSIP had an additional glycan at site 332, but this site is outside of the minimal V3

500
Bioinformatics analysis of sequencing data was performed using a zero-inflated generalized 501 Poisson significant-enrichment assignment algorithm to generate a -log10(p-value) for 502 enrichment of each phage clone across all samples, as previously described (24). Of note, the -503 log10(p-value) reproducibility threshold when testing these antibodies in PhIP-Seq was 2.3.

504
Thus, we considered a phage-displayed peptide as significantly enriched if its -log10(p-value) 505 was ≥ 2.3 in both technical replicates. A phage-displayed peptide was considered to be part of 506 the antibody's epitope sequence only if it was significantly enriched in both conditions tested (2 507 ng and 20 ng). Fold-enrichment of each phage-displayed peptide was also calculated across all 508 monoclonal antibodies tested.

510
Phage that were incubated without any monoclonal antibody served as a negative control for 511 non-specific binding of phage and were used to identify and eliminate background hits. For each 512 monoclonal antibody tested, enriched and unenriched peptides were aligned using Clustal 513 Omega. The minimal epitope of an antibody was defined as the shortest amino acid sequence 514 present in all of the enriched peptides. Logo plots were generated using WebLogo (PMID 515 15173120). For the "phage library" and "not enriched by nAbs" logo plots, only peptides that 516 spanned the full length of the minimal epitope (at least from S308 through D322) were included.

518
Analysis of RF-ADCC correlation with heterologous neutralization and V3 specificity 519 Unpaired, two-tailed t-tests were performed using GraphPad Prism 8.

521
Cell surface Env binding assays 522 Binding to cell surface-expressed Env was measured using a flow cytometry-based assay (42).

537
T332N was diluted to 1 µM in the same buffer as above and a series of four, two-fold dilutions of 538 Env trimer were tested as analyte in solution at a shake speed of 600 rpm at 30ºC. The kinetics 539 of mAb binding were measured as follows: association was monitored for 10 minutes, 540 dissociation was monitored for 6 minutes, and regeneration was performed in 10mM Glycine 541 HCl (pH 1.5). Binding-affinity constants (KD; on-rate, Kon; off-rate, Kdis) were calculated using 542 ForteBio's Data Analysis Software 7.0. Responses (nanometer shift) were calculated using data 543 that were background-subtracted from reference wells and processed by Savitzky-Golay 544 filtering, prior to fitting using a 1:1 model of binding kinetics.

570
Antibody repertoire sequence analysis and clonal family clustering 571 Sequences were preprocessed using FLASH, cutadapt, and FASTX-toolkit as previously 572 described (25, 43). Sequences from both technical replicates were combined, deduplicated, and 573 annotated with partis (https://github.com/psathyrella/partis) using default options including per-574 sample germline inference (44)(45)(46). Sequences with internal stop codons, or with out-of-frame 575 CDR3 regions were removed during this step. We did not exclude singletons in an attempt to 576 retain even very rare or undersampled sequences. Sequencing run statistics are detailed in 577 Table 1. Antibody sequences were merged with the 78 functionally-identified MG505 nAb heavy 578 and light chain sequences to form a single comprehensive MG505 P31 antibody sequence 579 dataset. This dataset was then used for clonal family analysis using both the partis unseeded 580 and seeded clustering methods (46). For the unseeded repertoire analysis, since we were 581 interested only in relative properties of clonal families, each data set was subsampled for 582 computational efficiency. For each dataset, three random subsamples of 50,000 sequences 583 were analyzed, comparing results among the three to ensure that they were large enough to 584 minimize statistical uncertainties. No subsampling was necessary for the seeded analysis.

586
Neutralizing antibody sequence analysis 587 Heavy and light chain nAb sequences ("seeds") were annotated, analyzed, and clustered into 588 clonal families with the MG505 P31 NGS sequences using the partis seeded clustering method 589 on the non-downsampled replicate-merged NGS dataset described above. NAb clusters were 590 delineated for Figure 2 using IgH chain variable region clustering information. Percent SHM was 591 calculated as the mutation frequency at the nucleotide level compared to the predicted naïve 592 allele, as determined by the per subject germline inference for MG505.

610
The funders had no role in study design, data collection and interpretation, or the decision to 611 submit the work for publication. 612 613